While the development of high-performance artificial intelligence working on cloud computer has improved convenience and affluence in people’s lives, its electric power consumption and a load of network has been increased as the amount of information flying through society has grown, and these issues will threaten our lives and the environment in the future. Although this problem could be greatly improved if acquired information data could be processed at an edge device before sending the data to the cloud computer, this scheme has not been implemented so far due to large computational cost raised by complex structure of used artificial neural network. However, it was found that reservoir computing has the advantages of low computing cost and fast processing speed due to its simple and small network structure. In particular, implementation of physical reservoir that is a physical device possessing distinctive properties required to reservoir layer (i.e., nonlinearity, short-term memory, and high dimensionality) is a promising approach to realize compact artificial intelligence integrated on the edge device (i.e., edge computer).[1,2] However, Even though the physical reservoir has been energetically attempted with various physical devices, some fatal issues still remain on the road to said implementation (e.g., low computational power, high electrical power consumption, and large device volume). Among such physical devices, although it was theoretically found that a physical reservoir utilizing nonlinear interfered spin wave can overcome these issues, the physical reservoir had not been experimentally demonstrated so far.[3] Very recently, we experimentally demonstrated a physical reservoir utilizing nonlinear interfered spin wave multi-detection for the first time [4,5] and found that this physical reservoir achieved excellent performance resulting from chaotic response, which has not been observed in the theoretical study. Here, we developed an innovative physical reservoir, namely opto-magnonic reservoir, based on the chaotic interfered spin waves through combining external field as well as the iono-magnonic reservoir [6] (Session 2.3-O1) to improve the computational performance of the physical reservoir based on nonlinear interfered spin waves. The opto-magnonic reservoir utilizes different interfered spin waves induced by visible light switching (i.e., dark or light) for reservoir computing. The visible light switching can manipulate spin configuration (i.e., high-spin configuration (ground state) or low-spin configuration (excitation state) in ferrimagnet Y₃Fe₅O₁₂ (YIG)), resulting in spin wave frequency manipulation. As a result, despite the overwhelmingly fewer number of node (100 nodes) than said iono-magnonic reservoir using 800 nodes, the processing error was drastically reduced to 4.96×10^-3 from 9.53×10^-3achieved with an iono-magnonic reservoir.[6,7]

In the presentation, we will report details of the opto-magnonic reservoir, including a mechanism of spin wave manipulation through the visible light irradiation. The reservoir computing powered by physical dynamics manipulation will be keys to implement the edge computer.